Back to articles
Volume: 54 | Article ID: art00007
Influence of the Number of Samples of the Training Set on Accuracy of Color Measurement and Spectral Reconstruction
  DOI :  10.2352/J.ImagingSci.Technol.2010.54.3.030501  Published OnlineMay 2010

In this article the authors analyze the influence of the number of samples in a training set on the accuracy of color and spectral measurements made using a colorimetric and multispectral imaging system. The authors develop a method for establishing the minimum and/or sufficient number of color samples in the training set above which the system's performance is independent of the number of samples. The authors also consider the dependence of the system's performance on the training set itself. Two setups of a charge coupled device camera-based imaging system are used for this purpose: a colorimetric configuration with three acquisition channels and a multispectral configuration with seven acquisition channels. On the basis of the criterion established in this article, the results show that the system's performance in terms of the accuracy of color measurement and spectral reconstruction seems to be independent of the training set used when over 110 samples are used for the colorimetric configuration and over 120 samples for the multispectral configuration. This result is true for both the number of samples in the training set and the training set itself. The method that the authors developed can be generalized and implemented in the industry for any application in which an imaging capture device is used for color and spectral measurements.

Subject Areas :
Views 43
Downloads 1
 articleview.views 43
 articleview.downloads 1
  Cite this article 

Marta de Lasarte, Montserrat Arjona, Meritxell Vilaseca, Jaume Pujol, "Influence of the Number of Samples of the Training Set on Accuracy of Color Measurement and Spectral Reconstructionin Journal of Imaging Science and Technology,  2010,  pp 30501-1 - 30501-10,

 Copy citation
  Copyright statement 
Copyright © Society for Imaging Science and Technology 2010
  Login or subscribe to view the content